LoginStart free

GPU profiling and observability

Understand what your GPU is doing.

Profile CUDA and ROCm workloads, inspect kernels and timelines, and monitor every GPU from one browser dashboard.

Start freeView demo

Open-source clientNo credit card

GPUFlight dashboard showing utilization, temperature, power, and activity across a GPU fleet

One place to profile, monitor, and learn

Start with the workflow you need today. The same dashboard connects the results.

Profile

Find the kernel that needs attention

Trace timings, occupancy, SASS, PC sampling, and memory behavior without losing the context of the full run.

Explore profiling

Monitor

See every GPU from one place

Follow utilization, memory, power, and temperature across hosts, then connect a system change to the workload behind it.

Explore monitoring

Build and learn

Run GPU code in your browser

Use Workbench for your own code or learn performance concepts through guided labs on real NVIDIA GPUs.

Open the playground

Follow the cause

From GPU activity to the kernel that caused it.

Start with the full run, open the kernel that stands out, and keep the evidence together instead of switching between disconnected captures.

  • Trace kernels, streams, and memory transfers on one timeline.
  • Drill into source, SASS, stall reasons, and measured occupancy.
  • Keep runs in one browser dashboard to compare and share.
See how profiling works
GPUFlight profiling session with merged passes, kernel metrics, and individual launch details

One command to your first profile

Use GPUFlight with an existing CUDA application, including PyTorch workloads.

gpufl trace --upload -- python train.py
  1. 1

    Run your workload

    Start with one command. Add deeper profiling passes only when you need them.

  2. 2

    Open the profile

    GPUFlight uploads the run and organizes kernels, timelines, memory activity, and metrics in your browser.

  3. 3

    Investigate and compare

    Find the bottleneck, share the session, or compare the result with another run or host.

Need SDK instrumentation or deployment options? Read the docs.

Choose what fits your workflow

Start free, add deeper practice tools, or scale profiling to more GPUs and people.

Free

$0

Explore GPUFlight and run occasional labs.

  • Monthly GPU credits
  • No daily run limit; runs use credits
  • 1 GB monthly ingestion
  • 7-day retention
  • Trace profiling
Start free

Practice

$6.99/month

Build hands-on CUDA performance skills.

  • Monthly GPU credits
  • No daily run limit; runs use credits
  • Four profiling passes
  • 1 GB monthly ingestion
Get started

Individual

$12.99/month

Profile and optimize your own workloads.

  • Monthly GPU credits
  • 5 GB monthly ingestion
  • 30-day retention
  • Up to 8 GPUs
Get started

Team

$49.99/month

Share GPU performance work with a small team.

  • Monthly GPU credits
  • 30 GB monthly ingestion
  • 90-day retention
  • 5 team members
Get started

Enterprise

Custom

Scale GPUFlight to your environment.

  • Custom GPU credits
  • Custom scale and retention
  • Security options
  • Direct support
Contact us

Every hosted GPU run uses credits from one monthly allowance shared between Performance Lab and Workbench. There is no daily run limit. Free includes RTX 6000 Ada access; H100 requires a paid plan. See your current credit allowance in the portal. The credit cost is shown before you run. CPU-only lessons and profiling your own hardware use no GPU credits.

Need a custom plan? Talk to us.

See what your GPU is actually doing.

Start with a free profile or explore the live demo.

Start freeView demo